An Optimization Study on Modular Reconfigurable Robots: Finding the Task-Optimal Design

Abstract

Reconfigurable collaborative robots are adaptive robotic systems offering new opportunities to rapidly create fit-to-task flexible automation lines demanded by the recent increasingly varying market needs in low-volume/high-mix industrial manufacturing. In this context, the configuration of a reconfigurable robot, comprising its morphology and its base placement, can be optimized for a certain criterion anytime the need for a change in batch production arises, boosting performances with respect to conventional fixed-structure robots. In this work we study and analyze how to find the best-suited reconfigurable robot topology for a given task, starting from a fixed set of joint and link modules. We make use of a two-stage generative design optimization approach to avoid running into issues from the “curse of dimensionality”. The efficacy of the approach is investigated and validated in a sequence of peg-in-hole tasks with a minimum-effort objective function. Simulation results are verified against real-world experiments with an in-house developed reconfigurable robot prototype comparing optimal against sub-optimal robot configurations.